2.2 Orthogonal Frequency Division Multiple Access based Cellular Network
2.2.3 Medium Access Control Layer
a system with more than one antenna at the transmitter and receiver, several data streams can be transmitted simultaneously. Thus, the system capacity can be increased tremendously by introducing the additional dimension of space now from another perspective [21], [22]. This dimension comes on top of time and frequency. In critical transmission situations transmit diversity can be exploited by transmission of redundant information at the antennas to make the transmission more robust. Spatial multiplexing receivers can be e.g. zero forcing (ZF), minimum mean squared error (MMSE) or successive interference cancellation (SIC) [23].
Massive MIMO uses an array of many antennas at the network side and a small number at the mobile device to form multiple beams to spatially reuse the available resources [24] [25].
With a huge number of subscribers, users have to be assembled in clusters, their movement has to be tracked and the interference of surrounding base stations has to be addressed. Using TDD mode, the channel can be estimated at the base station due to the reciprocity property.
Max rate The maximum rate scheduler is channel dependent as the resources are allocated according to a channel quality parameter. One condition could be the selection of the user with the highest SINR. In this way the time and frequency selective nature of the channel is exploited and the data rate maximized. On the other hand users with a low channel quality parameter are skipped or experience poor or even zero throughput and excessive delay.
Proportional fair The proportional fair scheduler provides a balance between scheduling at good channel conditions and fairness. This results in a higher cell throughput than the round robin technique but still maintaining commitment of the users experiencing poor channel conditions. The users are scheduled based on channel state information (CSI) measurements or channel quality indicator (CQI) reports. This results in a situation where users are scheduled on beneficial resources which provide good channel condi-tions. Thereby a high throughput is achieved. The utility value takes also into account the amount of allocated resources in the past, to maintain the fairness criteria, thus the users are scheduled when their instantaneous channel quality is high compared to its average. The time duration of one TTI is equal toTsf=1ms. The long-term average throughput is computed recursively as
ϒu(κ+1) =
βPFSϒu(κ) u∈/M(κ)
βPFSϒu(κ) + (1−βPFS)Ru(κ) u∈M(κ), (2.5) with the current aggregated throughputRu(κ)of useruduring TTIκ, forgetting factor βPFSfor windowing aspects, andMindicating the set of scheduled MS [26] [27]. The scheduling metric on each individual RB p for user u during TTI κ is the ratio of instantaneously possible and long-term throughput, with adjusting parameterαPFS for the fairness criteria:
MPFS,u,p(κ) = Ru,p(κ)
ϒαuPFS(κ). (2.6)
The user with the highest scheduling metric on RBpis eligible to transmit on this RB during TTIκ.
In conclusion, modern communication networks use channel dependent scheduling in order to encounter the special channel characteristics [28] and realize an optimal resource utiliza-tion.
An exemplary scheduling decision is shown in Figure 2.9, where the different users are separated by color. Dependent of the decision of the scheduler one user equipment (UE) also denoted as MS can have a different number of RBs assigned whereas each one can have a different MCS.
time QPSK, 16-QAM, or 64-QAM frequency subcarrier spacing = 15kHz
1 RB = 180 kHz = 12 subcarriers
1 slot= 0.5ms=
7 OFDMA symbols 1 subframe = 1ms = 1 TTI = 1 RB pair
UE 1 UE 2 UE 3
UE 4 UE 5 UE 6
Figure 2.9: Exemplary user assignment with time-frequency multiplexing of an OFDMAsignal (exemplary for normalCP) adapted from [6]
2.2.3.2 Hybrid Automatic Repeat Request
Channel decoding may fail and result in incorrect data, so erroneous data has to be detected and incorrect data discarded. Hence, in order to get more reliable data transmission the re-transmission protocol HARQ located on the MAC layer is introduced. HARQ combines channel coding and automatic repeat request (ARQ). When a UE detects an RB to be in-correctly received it can request a retransmission by sending negative acknowledge (NACK) on the UL. In FDD mode, with in total 8 HARQ processes, the message transmitted cor-responds to the DL packet received four subframes before [6]. In reaction, the identified erroneous data blocks are retransmitted. HARQ can be divided into a simple type 1 ver-sion where erroneous data is discarded and the retransmitted block is treated independently or a hybrid type 2 scheme, where the initial transmission is retained and combined with the retransmission [29] [11]. Type 2 includes also incremental redundancy, an adaptive error cor-rection technique. The ARQ protocol on the radio link control (RLC) on top of this scheme is introduced in order to account for missing data.
2.2.3.3 Channel Quality Parameters
In order to react to different channel situations the transmission of a channel quality report is essential. The data transmitted to and between base stations (BSs) is the channel state information (CSI). It comprises the following information:
Channel quality indicator (CQI) (Wideband or subband wide) The UE sends the infor-mation of the preferred MCS to the BS.
Precoding matrix indicator (PMI) Indicates the index within the codebook table which corresponds to the preferred precoding matrix determined at the receiver according to
a certain criterion based on the estimated channel. These are needed when precod-ing at the transmitter is used as for schemes as MU-MIMO and closed loop spatial multiplexing.
Rank indicator (RI) (Wideband) Gives the number of useful transmission layers for spa-tial multiplexing.
The reporting of the channel feedback information as CSI may be periodic or aperiodic.
2.2.3.4 Link Adaptation
Link Adaptation adjusts the transmission parameters for the individual users to the present channel conditions. The desirable MCS is selected in order to maintain QoS criterias while realizing high throughput rates. Dependent on the signal-to-noise ratio (SNR) on the individ-ual RBs a set of modulation scheme and code rate is selected. An overview on the different modulation schemes is presented in Section 2.1.1. Higher order modulations provide higher data rates and are used when a good link with adequately high SNR is available. In case of inferior channels, the MBS decides on a more robust lower modulation scheme as QPSK due to its fewer sensitivity to interference, noise, and estimation errors of the channel. The available coding schemes for the particular modulation schemes are{1/9, 1/6, 0.21, 1/4, 1/3, 0.42, 1/2, 0.58, 2/3, 0.73}for QPSK,{0.43, 0.46, 1/2, 0.54, 0.58, 0.61, 2/3, 0.73, 4/5}for 16-QAM, and{0.58, 0.62, 2/3, 0.70, 0.74, 4/5, 0.85, 0.9}for 64-QAM [30]. In order to address the time variation of the channel, especially critical for uplink transmission, an additional outer loop link adaptation is introduced [31].
2.2.3.5 Link to System Mapping
In order to encounter the huge amount of data to be simulated, an abstraction level, including physical as well as system level aspects, has to be introduced. To address the complex and computationally complex simulations a link-to-system (L2S) level interface connects the two levels. The system level contains all connections between all kinds of BSs to the MSs and vice versa, whereas the link level considers the link-pair of transmitter and receiver.
Measures on the link level are bit error rate (BER) and block error rate (BLER), whereas on the system level they are throughput of the cells and the system [32]. The L2S-level simulation mapping is performed through the use of look-up-tables. The tables are gained by performing link-level simulations. They contain the information on the BLER as a function of the SINR. The SINR of each subcarrier is determined. Subsequently, the value of the individual subcarriers is mapped to an effective SINR whereof the corresponding BLER can be gained. Figure 2.10 depicts the procedure of the mapping between link and system level according to [33].
In the literature, several methods which take into account the capacity effective, exponen-tial effective, logarithmic, and mutual information effective SINR are presented [32]. An
attractive method for the L2S-level mapping is the mutual information effective SINR map-ping (MIESM) as it supports a mapmap-ping with different modulation and coding schemes and due to its high accuracy.
Link Level
BLERAWGN
(PHY Abstrac-tion Mapping)
Mapping Function as MIESM, EESM
System Level
Link adaptation, scheduling, ARQ etc.
• Generate frequency selective channelH(f)
• Determine received SINR of each sub-carrier
BLER
γ¯ of each subchannel
Throughput, packet error rate etc.
Figure 2.10: Procedure of the link-to-system-level mapping
The effective SINRγeffis calculated based on a non-linear mapping function with the use of the mutual information functionIbmwhich considers the specific modulation,bmthe complex QAM-symbol, the referencebref as the average number of transmitted bits per resource, a calibration factorβ,γmthe SINR of subcarrierm, and the number of subcarriers per transport blockNblocks[33]:
γeff=βIb−1
ref
1 Nblocks
Nblocks
X
m=1
Ibm γm
β
!
(2.7)
The calibration factorβis chosen to minimize the root mean square error betweenγeffand the static SNR leading to the same BLER. The mutual information depends on the modulation alphabet size 2bm (withbm bit per QAM symbol) and an expectation term. This results in curves which map the SNR for the individual MCS to the mutual information. The SNR is defined as the ratio of mean QAM symbol energy Es to noise power densityN0. While for low SNR they follow the Shannon capacity, they saturate atbmin the high SNR regime. The saturation due to the limited QAM symbol alphabet is shown in Figure 2.11 whereas Gray labeling is assumed [32] [34].
SNR=NEs
0
Channelcapacityinbitpersymbol
256-QAM 64-QAM 16-QAM Shannon bound
Figure 2.11: Capacity of complex-valued AWGN channel with different QAM input constellations [34]
2.2.3.6 Coordinated Multipoint
Coordinated multipoint (CoMP) is a method for cooperation between cells of homoge-neous and heterogehomoge-neous deployments in order to improve the network performance [35].
Geographically separated base stations transmit and receive data in a coordinated way. It comprises joint transmission and reception as well as inter-cell coordinated scheduling. A central network element coordinates the radio frequency transmission at the spread antenna locations [12]. Especially the data rates for the cell edge users can be increased due to the reduced interference situation.
2.2.3.7 Handover
While some users of a cellular system might reside at one location others move with pedes-trian speed or travel in cars and highspeed trains. Due to the limited coverage of one cell a handover procedure has to be integrated. In case the signal from the neighboring cell is higher than the one from the associated cell a seamless handover is favored in order to con-stantly maintain the QoS. A handover can be controlled by the network as well as requested by the UE. In case the serving BS and the requested BS are directly connected by the logi-cal X2-interface, the handover decisions are negotiated between these two cells. Otherwise, the EPC is involved in the handover procedure. Classical handover scenarios are on the macrocell layer from one MBS to another. With the introduction of an additional layer, fur-ther unsymmetrical inter-layer handover scenarios such as from MBS to HBS and HBS to MBS as well as inter femtocell-layer handovers from HBS to HBS come in addition. Within these multi-layer scenarios further aspects, as the unsymmetrical power levels, increasing handover rates as well as the admission control due to limitations such as closed subscriber group (CSG), have to be addressed.
2.2.3.8 Carrier Aggregation
To reach the requirements of the radio communication sector of the International Telecom-munication Union (ITU) the concept of carrier aggregation (CA) was introduced. To guaran-tee backward compatibility it has been decided not to increase the overall system bandwidth of maximum 20MHz but to add several channels intra-band (consecutively or discontinu-ously) and also inter-band at different frequency ranges and aggregate them. With this also fragmented spectrum can be exploited. One individual channel is denoted as a component carrier (CC). With CA the usable bandwidth can be increased dependent on the LTE Re-lease from five CC in ReRe-lease 12 up to 32 CC in ReRe-lease 13 by aggregating multiple carrier components. This results in a maximum available bandwidth of 100MHz in Release 12 to 640MHz in Release 13 [12] [36]. Figure 2.12 depicts an example of FDD carrier aggregation where legacy Release 8 and 9 UEs get only one of the component carriers and LTE-A UEs can transmit on several CCs simultaneously. The UEs from previous releases are denoted by R8/R9 UE. Their restriction on only one CC is illustrated with their blue and red color. The blue colored UE is only eligible to use CC 1 printed with the same color and the red colored UE transmits on CC three. Different an LTE-A UE can use several up to all available CCs.
The aggregation of CCs can be different for DL and UL transmission. As the CCs can be adjacent or spread over the frequency, different scenarios have to be discussed. In case they are co-located they span similar coverage areas. Contrary, when the CCs are non-contiguous in different frequency band areas the coverage area can differ significantly which has to be addressed. Mobility and handover aspects need to be based on the CC with lower frequency, as this results in a wider coverage area due to lower path loss.
R8/R9 UE
R8/R9 UE LTE-Advanced UE
max 5 CC, max 100 MHz Component Carrier, CC
Same DL and UL allocation Different DL and UL allocation CC BW: 1.4,3,5,10,15,20 MHz
fDL
fDL fDL
fUL
fUL fUL
MBS
Figure 2.12:Overview on carrier aggregation adapted from [36]
Carrier aggregation can be also thought of together with different beams of sectorization and separate component carriers for macrocells and small cells. Nevertheless, this reduces the overall available resources when pre-assigning the bandwidth completely.